Monday, November 19, 2012

"Evidence" swamped by human foibles -- again

Here we are in the age of "evidence-based medicine," when medical practice is supposed to be standardized and based on empirical results.  So how is the kind of story in yesterday's New York Times still possible?  In 'How Back Pain Turned Ugly' Elizabeth Rosenthal tells how a batch of injectable steroids destined to treat chronic lower back pain ended up killing 32 people and injuring 438.  And, even on a good day it's not clear how many this treatment would have helped.

There are numerous aspects of this story that just shouldn't have happened.  Of course, the malfeasance in the factory, The New England Compounding Center, is number one.  They were lax in their quality oversight, so much so, Rosenthal writes, that some of the steroid vials they shipped to clinics were so contaminated that a visible white fuzz was floating in them.  If this is true, this is a tragic example of people only seeing what they expect to see.  The evidence that this was a tainted drug was right before the eyes of the medical personnel who took the vials out of their refrigerators and filled syringes with the contaminated stuff, but they didn't see it.  Too often, what constitutes 'evidence' is often what we expect to see.


This picture of a hamburger, for example, is shades of red and grey, but if you aren't color blind you are seeing a red tomato, green lettuce, yellow cheese and a beige-ish bun.  But,  those colors aren't actually on your screen.  Your brain is filling them in.  This is a point made by Edwin Land, founder of Polaroid film technology.

In the clinic, a doctor will see sterile medicine when that's what she or he expects.  (Yes, experimental results can get interpreted in this way, too -- as can projected election winners, and, say, whether Fox News should have called Ohio for Obama so early in the evening of Nov 6.)

A second interesting aspect of the back pain story is that how to treat back pain is, speaking of shades of grey, still a grey area in medicine.  When should it be treated, and with surgery or with steroid injections?  Doctors pretty much agree that the pain should have lasted 4-6 weeks before any treatment.  Evidence?  As Rosenthal says,
... steroid shots are not a cure-all, even for the conditions for which doctors agree an attempt is worthwhile: low back pain accompanied by signs of nerve injury like tingling or weakness in a leg. One-third of such patients will get better, one-third will show some improvement and some will show no improvement at all...
This is just the kind of situation we write about all the time here on MT with respect to prediction of disease from genes; just as every genome is unique, every case of lower back pain is unique because every person has a different history of back injury, stresses, muscle strength, pain tolerance, and so forth, making it very difficult if not impossible to predict who will respond well to which treatment.  Thus, the probability of improvement is 1/3, but the probability of getting worse is also 1/3.  Though, the probabilities are likely to be about the same with no treatment at all.

Similarly, genetics wrestles with the degree to which a given genotype predicts 'the' trait and trying to specify some probability of that.  But if the trait is itself variable, perhaps the issue would be to predict, say, a range of severity or manifestation.  But this is a very slippery often perhaps self-confirming way to think, given that we can imagine symptoms if we expect them, or can force a diagnosis we want to see.

And this is just like trying to predict who will have a heart attack, a stroke, develop type 2 diabetes, or Alzheimer's disease, based on the best evidence we have -- genetic indicators and lifestyle risk factors.  That is, we're not yet very good at it except when very strong risk factors are involved.  Indeed, even one of the strongest environmental risk factors we know of, smoking, isn't all that useful in predicting on an individual level who will or won't get sick from it.

"Evidence-based" assessment and treatment of back pain shares something else with genetics.  And that is that it's equally confounded by, entangled with and driven by financial or other material or career or professional kinds of interests considerations. As Rosenthal writes:
The shots — which may include a steroid and an anesthetic — are often dispensed at for-profit pain clinics owned by the physicians holding the needle. “There’s a lot of concern about perverse financial incentive,” Dr. Friedly added.
The increase in treatment has not led to less pain over all, researchers say, and is a huge expense at a time of runaway health costs. “There are lots of places doing lots of injections for conditions that haven’t been shown to benefit,” says Dr. Janna Friedly, a researcher at the University of Washington, who added, “Sadly, some of the patients who got meningitis were probably in that category — they did not have conditions where steroid injections were indicated.”
And when that's true, it takes a lot of evidence to show that a technology or treatment that a lot of people have invested in really does not work, and shouldn't be done.  Sounds rather reminiscent of direct-to-consumer genetic prediction companies to us.  Indeed, Rosenthal says that "studies are at best inconclusive about exactly which groups of back-pain patients are likely to benefit from steroid shots."  Studies. Evidence. Murky conclusions.  

What does the evidence unequivocally show?  That medical practice is as equally confounded by human foibles, including greed, as is genetic research, and that complexity makes evidence very hard to interpret.

No comments: